Issue description:
SchedulerDAGInstrs::buildSchedGraph ignores dependencies between FixedStack
objects and byval parameters. So loading byval parameters from stack may be
inserted *before* it will be stored, since these operations are treated as
independent.
Fix:
Currently ARMTargetLowering::LowerFormalArguments saves byval registers with
FixedStack MachinePointerInfo. To fix the problem we need to store byval
registers with MachinePointerInfo referenced to first the "byval" parameter.
Also commit adds two new fields to the InputArg structure: Function's argument
index and InputArg's part offset in bytes relative to the start position of
Function's argument. E.g.: If function's argument is 128 bit width and it was
splitted onto 32 bit regs, then we got 4 InputArg structs with same arg index,
but different offset values.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@165616 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index d4d29f5..e6f0ef2 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -2534,6 +2534,8 @@
void
ARMTargetLowering::VarArgStyleRegisters(CCState &CCInfo, SelectionDAG &DAG,
DebugLoc dl, SDValue &Chain,
+ const Value *OrigArg,
+ unsigned OffsetFromOrigArg,
unsigned ArgOffset) const {
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo *MFI = MF.getFrameInfo();
@@ -2561,7 +2563,7 @@
getPointerTy());
SmallVector<SDValue, 4> MemOps;
- for (; firstRegToSaveIndex < 4; ++firstRegToSaveIndex) {
+ for (unsigned i = 0; firstRegToSaveIndex < 4; ++firstRegToSaveIndex, ++i) {
const TargetRegisterClass *RC;
if (AFI->isThumb1OnlyFunction())
RC = &ARM::tGPRRegClass;
@@ -2572,7 +2574,7 @@
SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i32);
SDValue Store =
DAG.getStore(Val.getValue(1), dl, Val, FIN,
- MachinePointerInfo::getFixedStack(AFI->getVarArgsFrameIndex()),
+ MachinePointerInfo(OrigArg, OffsetFromOrigArg + 4*i),
false, false, 0);
MemOps.push_back(Store);
FIN = DAG.getNode(ISD::ADD, dl, getPointerTy(), FIN,
@@ -2606,14 +2608,16 @@
CCInfo.AnalyzeFormalArguments(Ins,
CCAssignFnForNode(CallConv, /* Return*/ false,
isVarArg));
-
+
SmallVector<SDValue, 16> ArgValues;
int lastInsIndex = -1;
-
SDValue ArgValue;
+ Function::const_arg_iterator CurOrigArg = MF.getFunction()->arg_begin();
+ unsigned CurArgIdx = 0;
for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
CCValAssign &VA = ArgLocs[i];
-
+ std::advance(CurOrigArg, Ins[VA.getValNo()].OrigArgIndex - CurArgIdx);
+ CurArgIdx = Ins[VA.getValNo()].OrigArgIndex;
// Arguments stored in registers.
if (VA.isRegLoc()) {
EVT RegVT = VA.getLocVT();
@@ -2709,7 +2713,8 @@
if (Flags.isByVal()) {
unsigned VARegSize, VARegSaveSize;
computeRegArea(CCInfo, MF, VARegSize, VARegSaveSize);
- VarArgStyleRegisters(CCInfo, DAG, dl, Chain, 0);
+ VarArgStyleRegisters(CCInfo, DAG,
+ dl, Chain, CurOrigArg, Ins[VA.getValNo()].PartOffset, 0);
unsigned Bytes = Flags.getByValSize() - VARegSize;
if (Bytes == 0) Bytes = 1; // Don't create zero-sized stack objects.
int FI = MFI->CreateFixedObject(Bytes,
@@ -2732,7 +2737,8 @@
// varargs
if (isVarArg)
- VarArgStyleRegisters(CCInfo, DAG, dl, Chain, CCInfo.getNextStackOffset());
+ VarArgStyleRegisters(CCInfo, DAG, dl, Chain, 0, 0,
+ CCInfo.getNextStackOffset());
return Chain;
}
diff --git a/lib/Target/ARM/ARMISelLowering.h b/lib/Target/ARM/ARMISelLowering.h
index 2b8f382..b5020c9 100644
--- a/lib/Target/ARM/ARMISelLowering.h
+++ b/lib/Target/ARM/ARMISelLowering.h
@@ -466,7 +466,10 @@
SmallVectorImpl<SDValue> &InVals) const;
void VarArgStyleRegisters(CCState &CCInfo, SelectionDAG &DAG,
- DebugLoc dl, SDValue &Chain, unsigned ArgOffset)
+ DebugLoc dl, SDValue &Chain,
+ const Value *OrigArg,
+ unsigned OffsetFromOrigArg,
+ unsigned ArgOffset)
const;
void computeRegArea(CCState &CCInfo, MachineFunction &MF,